Skip to content
Crypto.world

The Ledger

Ethereum Draft Sets Out Post-Quantum Keys for Validators

Draft EIP-8450 proposes an Ethereum registry for validators to register hash-based keys before a later switch away from BLS signatures. The proposal sets an eight-registration block limit and leaves the new keys unused until a future fork.

Ethereum Draft Sets Out Post-Quantum Keys for Validators

“The registered keys are dormant: no duty is authenticated with them in this fork.”

Thomas Coratger, Kevaundray Wedderburn, Tom Wambsgans and Justin Drake, authors of EIP-8450

EIP-8450 separates registration from the switch

Draft EIP-8450 proposes a registry in Ethereum’s beacon state where each validator can register a post-quantum public key ahead of a later change to the network’s signature scheme. The authors created the proposal on October 15, 2026. It remains a draft, and it does not change how validators authenticate their duties: BLS signatures stay in use until a separate fork switches those duties to the registered keys.

The proposed key scheme, leanXMSS, uses hash-based signatures built on Keccak-256. Each validator would register one public key tied to its existing validator index, along with a proof that it controls the key. The registry would let the network measure what share of stake has registered keys before deciding when to undertake a signature change.

The authors frame early registration as preparation and a possible recovery anchor. If a cryptographically relevant quantum computer breaks BLS before the later switch, validators with keys registered beforehand would have an alternative that does not depend on BLS signatures made after the break. The proposal does not claim that such a computer exists or set a date for the switch.

BLS operators and withdrawal addresses get different powers

EIP-8450 gives validators two registration paths. An operator can use the validator’s current BLS signing key to register a key, but cannot use that path to replace one already registered. The validator’s withdrawal address can authorize a registration or replacement through a new execution-layer request tied to EIP-7685.

The distinction addresses the risk that a quantum attacker could forge BLS signatures. If BLS becomes vulnerable, the operator path would still permit registrations for validators that have not registered, but it could not overwrite their existing keys. The withdrawal-address path is the only route to replacement, though its security depends on the withdrawal address itself; an externally owned address secured by ECDSA would face a related quantum risk.

Under the proposal, a withdrawal address authorizes a commitment to the key on the execution layer, while the full key and proof arrive through a consensus-layer registration. An authorization expires after 256 epochs. The draft also caps registrations at eight per beacon block and sets a separate maximum of 16 authorization requests per execution payload.

The two-layer key trades larger signatures for smaller setup steps

The leanXMSS design has two tree layers, each with height 16. Together they support 232 one-time signature indices, while limiting each key-generation step to 216 one-time keys. Signers must track which indices they have used: the same index cannot sign two different messages, or the key can become forgeable.

The draft compares its two-layer design with a flat tree of height 32. It gives a two-layer signature size of 2,404 bytes, against 1,460 bytes for the flat version, and estimates 186 Keccak-f permutations for verification rather than 109. The added signature data reduces the scale of key generation: the proposal says a million validators could register in about 125,000 full blocks, or roughly 17 days at 12-second slots, if every block reached the eight-registration limit.

Each registration is 2,541 bytes, including its proof of possession. At the block limit, the draft estimates about 20 kilobytes of additional data and 1,488 Keccak-f permutations per block. Those limits bound the work imposed on blocks, but do not guarantee that validators will complete registration on schedule.

A registered key is not proof of operational readiness

The proposal says a valid proof establishes that the key holder could sign at registration time; it does not establish that a validator will keep its seed, preserve its signing state, or run compatible software when duties eventually switch. The authors therefore describe registered stake as an upper bound on readiness, not a count of validators prepared to operate under the new scheme.

EIP-8450 requires changes to both Ethereum’s consensus and execution layers. Its text says the leanXMSS scheme has a specification and implementation, while the consensus-layer changes and request contract still need to be provided before the EIP moves beyond draft status. The proposed registry would store one 33-byte entry per validator; at one million validators, the draft estimates roughly 33 megabytes of registry data.

Sources

  1. Original source Ethereum Improvement Proposals eips.ethereum.org

Related stories

Explore

More articles